Identifying the proteins to which small-molecule probes and drugs bind in cells Shao-En Onga,1, Monica Schenonea, Adam A. Margolinb, Xiaoyu Lic, Kathy Doa, Mary K. Doudd, D. R. Mania,b, Letian Kuaie, Xiang Wangd, John L. Woodf, Nicola J. Tollidayc, Angela N. Koehlerd, Lisa A. Marcaurellec, Todd R. Golubb, Robert J. Gouldd, Stuart L. Schreiberd,1, and Steven A. Carra,1 aProteomics Platform, bCancer Biology Program, cChemical Biology Platform, dChemical Biology Program, and eStanley Center for Psychiatric Research, The Broad Institute of MIT and Harvard, 7 Cambridge Center, Cambridge, MA 02142; and fChemistry Department, Colorado State University, Fort Collins, CO 80523 Contributed by Stuart L. Schreiber, January 15, 2009 (sent for review December 21, 2008) Most small-molecule probes and drugs alter cell circuitry by interact- known (the ‘‘target I.D. problem’’). It could provide strong clues to ing with 1 or more proteins. A complete understanding of the the mechanisms used by SMs to achieve their recognized actions interacting proteins and their associated protein complexes, whether and it could suggest potential unrecognized actions. the compounds are discovered by cell-based phenotypic or target- Strategies for ‘‘target identification’’ have been developed that based screens, is extremely rare. Such a capability is expected to be rely on genetic (9), computational (10, 11) and biochemical (12) highly illuminating—providing strong clues to the mechanisms used principles. Although several key molecular targets have been iden- by small-molecules to achieve their recognized actions and suggest- tified through affinity chromatography (13–15), it has not been ing potential unrecognized actions. We describe a powerful method widely applied as a general solution to target identification for a combining quantitative proteomics (SILAC) with affinity enrichment number of reasons. It is often challenging to prepare SM affinity to provide unbiased, robust and comprehensive identification of the reagents that retain the desired cellular activity. Experiments with proteins that bind to small-molecule probes and drugs. The method SM baits, even more so than antibody-based immunoaffinity re- is scalable and general, requiring little optimization across different agents, require carefully chosen and effective controls as baits may compound classes, and has already had a transformative effect on our vary considerably in their chemical structures and binding proper- studies of small-molecule probes. Here, we describe in full detail the ties. Moreover, high stringency washes are required to minimize SCIENCES application of the method to identify targets of kinase inhibitors and contamination associated with nonspecific, and bait-independent, APPLIED BIOLOGICAL immunophilin binders. interactions of cellular proteins with the reagents. The latter shortcoming is especially significant as it biases toward high-affinity SILAC ͉ small molecules ͉ target identification interactions, decreasing the likelihood of identifying more weakly bound proteins or protein complexes that may play significant roles any small-molecule (SM) probe or drug discovery efforts in the polypharmacology of a SM. Mstart by selecting a target that is expected to modulate a Classically, identifying targets of SMs through biochemical pu- pathway or disease of interest. Some drug-discovery efforts opti- rification relied on large amounts of starting protein, extensive mize existing compounds so that they bind their intended targets protein fractionation, stringent wash conditions, gel visualization with higher specificity and affinity. By focusing on specific protein and excision of specific bands to yield only the most directly and classes (e.g., kinases), this paradigm of drug discovery routinely uses tightly bound proteins (13, 16, 17). With proteomic MS approaches in vitro assays with recombinant proteins in binding or biochemical (18), even affinity pull-down experiments generate large protein assays (refs. 1 and 2; also recently reviewed in ref. 3). Although such catalogs, inflating the list of candidate ‘‘hits’’ and requiring, some- large-scale screens can provide early leads that perform well against times arbitrary, prioritization of these proteins for validation. a specific target, the absence of a biological context results in higher Quantitative proteomics has proven to be a powerful tool for attrition rates in later stages of drug development arising from discriminating specific protein–protein interactions from back- unanticipated or undetected off-target effects, or lack of relevance ground interactions in affinity pull-downs (19, 20). Although this of the target protein to the underlying disease process. Further- was recently applied to profile kinases enriched in kinase inhibitor more, screens using purified protein substrates do not accurately pull-downs (21, 22), these experiments still assumed kinases as represent biological levels of target proteins, potentially leading to targets a priori and did not use quantitative data to define SM generation of incorrect hypotheses for on- or off-target drug effects. specific targets. From the standpoint of drug safety and efficacy, unbiased identi- Here, we describe an analytical framework combining quantita- fication of proteins and associated molecular complexes that bind tive mass spectrometry (MS)-based proteomics (23) with affinity to a drug allows direct evaluation of its polypharmacology (4) and chromatography for unbiased, sensitive, specific and comprehen- provides valuable insight into its mode of action and avenues for sive determination of SM-protein interactions within cellular pro- compound optimization. teomes. We use SILAC to distinguish cell populations for our SM Cell-based phenotypic screens allow the discovery of compounds affinity enrichments (23). Cells are cultured in growth medium that induce state transitions in cells or organisms without bias containing either ‘‘light,’’ natural isotope abundance forms, or the regarding specific targets, pathways or even processes. This discov- ‘‘heavy,’’ 13C, 15N-bearing versions of arginine and lysine. Growing ery-based approach has been used with increasing frequency and success in recent years (5, 6). The ability to define cell states globally and molecularly in the context of high-throughput screens, for Author contributions: S.-E.O., X.L., N.J.T., A.N.K., and L.A.M. designed research; S.-E.O., M.S., X.L., K.D., M.K.D., and L.A.M. performed research; S.-E.O., A.A.M., D.R.M., L.K., X.W., and J.L.W. example, using imaged cellular features (7) and mRNA expression contributed new reagents/analytic tools; S.-E.O., M.S., A.A.M., K.D., M.K.D., T.R.G., R.J.G., S.L.S., (8), suggests that its impact will continue to grow in the future. and S.A.C. analyzed data; and S.-E.O., T.R.G., R.J.G., S.L.S., and S.A.C. wrote the paper. However, as with SMs emerging from target-based screens, there The authors declare no conflict of interest. exists currently no reliable way to assess the complete set of proteins Freely available online through the PNAS open access option. that interact with SMs discovered in phenotype-based screens. Such 1To whom correspondence may be addressed. E-mail: [email protected], a capability is expected to be highly illuminating. It is especially stuart࿝[email protected], or [email protected]. critical with SMs identified in phenotype-based screens because This article contains supporting information online at www.pnas.org/cgi/content/full/ even the target relevant to the induced phenotype is usually not 0900191106/DCSupplemental. www.pnas.org͞cgi͞doi͞10.1073͞pnas.0900191106 PNAS Early Edition ͉ 1of6 Downloaded by guest on September 30, 2021 Fig. 1. Identifying specific SM-protein interactions with quantitative proteomics. (A) SILAC identifies specific ABFK506-binding protein 4 Histone H1.3 protein interactions with SM baits. Cell populations are S AEASSGDHPTDTEMK ASGPPVSELITK Specific Non-specific Light SM Intensity -Beads Digest S fully labeled with light (black) and heavy amino acids Combine, Identify NS protein Intensity wash beads, and Intensity with (red) and lysates incubated either with SM-loaded beads boil beads, quantify trypsin NS (SM-Beads) and soluble SM competitor or SM-Beads SDS-PAGE by MS alone. Proteins interacting directly with the SM or via m/z Heavy SM-Beads Legend secondary and/or higher order interactions (marked ‘‘S’’ Small molecule 524 526 528 530 532 598 600 602 604 606 for specific) will be enriched in the heavy state over the Bead m/z m/z Proteins light and will be identified with differential ratios. Non- specific (NS) interactions of proteins will be enriched equally in both states and have ratios close to 1. (B) Experimental mass spectra showing specific protein interactions with the immunophilin ligand, AP1497. (Left) A peptide from FKBP4, a known binding partner to FK506, is observed with a highly differential ratio. (Right) In contrast, a histone H1.3 peptide is identified with a ratio close to 1, indicating no specific binding to the soluble SM competitor. and dividing cells incorporate these amino acids in their proteomes, the importance of appropriate controls along with highly accurate reaching full incorporation after 5 population doublings and pro- quantitative measures and statistical methods that allow the dis- 13 ducing the characteristic mass shift, 6 Da with C6-Argor8Dain crimination of real SM-protein interactions from nonspecific bind- 13 15 C6 N2-Lys containing peptides, observable by MS. Importantly,